What You'll Learn

Agents can run a hundred analyses before lunch. That makes it easy to believe things, and hard to know which ones you should. Vibe Research is four four-hour sessions where you and your agent take one question through a whole research cycle: what's already known, a design that could fail, reading what came back, and a gate built from your own mistakes. You'll be handed analyses that look careful and are wrong in one specific way, and you'll find what's wrong. Every session ends with a mistake you caught, written so someone else could catch it too. By the last session those mistakes are your gate, and you'll stand in front of the room and say what you believe, why, and what would change your mind.

- Predict, then look: write down what you expect before your agent runs anything, then score yourself honestly. - Check a number against the thing that produced it, not the text that quotes it. - Design a test that could come back against you, and know what result would make you believe your own claim. - Read a result for what it actually measured, and test an explanation before you accept it or throw it away. - Build a review gate from your own failures, prove it can fire, and split the checking between you and your agent. - Defend a finding in front of the room: what you believe, why, and what would change your mind.

Course Schedule

4 Weeks
18 Classroom Hours

Week 1 1 session

Session 1: The question

Thursday, November 12 at 10:00 AM - 2:30 PM PST

Ask a question that could turn out false, survey what's already known with sources you can open yourself, and seal your first prediction before anyone looks. Homework: draft a two-condition design for session 2 (a four-line template is posted).

4.5h

Week 2 1 session

Session 2: Designing something that can fail

Thursday, November 19 at 10:00 AM - 2:30 PM PST

Find the difference that got there before your variable did, choose a control that is capable of failing, and pre-register a test that your agent then tries to break. Homework: finish the pre-registration, then run the smallest pilot that could produce a signal.

4.5h

Week 9 1 session

Session 3: Reading what came back

Thursday, January 07 at 10:00 AM - 2:30 PM PST

Four ways a result can mislead you, one an hour, each from a published case. Then write the limits section first, and say what you now believe.

4.5h

Week 11 1 session

Session 4: The gate and the room

Thursday, January 21 at 10:00 AM - 2:30 PM PST

Turn the mistakes you caught into a gate that runs, prove it can fire, and present one finding to the room: what you believe, why, and what would change your mind.

4.5h

All times shown in Pacific Time

Who Is This Course For

People who build with agents and want to find things out with them

Who this is for & what you'll need

🟡 Guided codingYou'll run commands and edit files. No prior experience needed.
🖥️ A few installsYou'll install a couple of tools — we'll help you in class.
On your own machineAny agent you built in Intro to Agents that can search the web and show its sources
Or use the hosted optionClaude Code with a Claude subscription or API access or similiar frontier model: the setup from Agentic SDLC

Frequently Asked Questions

How are classes delivered?

All classes are live and online via video call. Sessions are recorded so you can review them later if you miss one.

What does "pay what you can" mean?

We show a suggested price, but you can pay any amount. If you need a larger discount, apply for a scholarship — we offer 100% scholarships and no one is turned away for lack of funds.

Do I need prior experience?

🟡 Guided coding. You'll run commands and edit files. No prior experience needed.

What do I get when I enroll?

Live sessions with the instructor, access to course curriculum materials, community access via our Matrix chat server, and recordings of all sessions.

Want to schedule this course for your team?

Contact us: [email protected]